What are AI Overviews?

AI has been integral to Google’s ecosystem for over a decade. More recently, with the proliferation of generative AI, Google is integrating this innovative technology into multiple products across its product line. In search, in the second half of 2023 Google embarked on a year-long experiment in labs called Search Generative Experiences (SGE). During that time Google tested and learned how generative AI could be most effective for search. We monitored that activity with BrightEdge Generative Parser™ (BGP) which we published in our Ultimate Guide to Google SGE.

On May 14th 2024, Google announced that Search Generative Experiences (SGE) would become AI Overviews (AIO) and start appearing for users outside of labs. AI Overviews are built on the same Google Gemini AI Language Model as SGE, and BGP can track their presence just as it did with SGE. AI Overviews are available to US users who are logged in to Google, regardless of whether they’ve enrolled in Google Labs. However, as of this writing, they are still not available to non-logged-in users. This development represents a significant step forward for AI-powered search and has major implications for search engine optimization (SEO). Now, SEO must account for traditional organic search results as well as the citations that AI Overviews use to answer questions.

At BrightEdge, our mission is to help marketers understand how these innovations from search engines will impact how they connect with their customers. This guide serves as your resource for AI Overviews, equipping you with the knowledge to understand how generative AI technology affects how your audience interacts with digital content.

 

How do AI Overviews appear?

Presently, AI Overviews are limited to the US and appear for users who are logged into Google. However, Google has confirmed it is showing AI Overviews to users who are not logged in during small tests.
AI Overviews consist of short paragraphs, bulleted lists, images, and product carousels that Google generates using Generative AI to add context and summarize search results. They appear at the top of the search results, below most paid and above organic listings. To date, AI Overviews appear in a collapsed state, requiring the user to click to read the full content:

When the user clicks "Show More", they are taken to an experience where different modules provide greater detail on the answer. Citations for where the AI Overview obtained the information to compose its content are also shown throughout the experience. In this case, we see that TYR, RunRepeat and the American Academy of Podiatric Sports are cited to for the Generative AI’s content:

The degree to which AI Overviews appear for a user can be controlled (although not entirely turned off) in the Labs section of Google. By enabling AI Overviews, it will deliver fuller, more comprehensive answers like the one above. Google has confirmed that turning this feature off will not eliminate AI Overviews from the experience entirely.

For users in the US who are logged in, AI Overviews are currently appearing in less than 15% of all queries. For marketers who need to understand how AI Overviews are appearing for their keywords, it is important to ensure you’ve enabled these results in Google Labs to see the most comprehensive version that your audience may see. Based on Google’s deployment, it is also crucial to identify which keywords are most likely to trigger an AIO.

 

The presence of AI Overviews varies based on the industry. Data suggests this correlation is largely due to the nature of the industry. For example, in healthcare or B2B technology, where queries are more informational in nature, there is a significant difference in presence compared to industries like restaurants or travel, where user behavior is more transactional (booking reservations, finding addresses, calling the restaurant). This observation holds in other industries as well, except for finance, where Google is not deploying AI Overviews frequently.

 

When do AI Overviews (AIO) Appear?

AI Overviews are largely appearing for informational queries. Fortunately for marketers, there are signals from traditional search features that can be used as a proxy to estimate where an AIO is likely to appear. According to BrightEdge Generative Parser™, there are correlations between SERP features and the presence of AIO.

As displayed here, featured snippets have the strongest correlation with the presence of an AI Overview. This makes sense because featured snippets mostly appear for informational queries. Conversely, certain navigational and transactional elements have a negative correlation with the presence of AI Overviews. Local 3 Packs, which are deployed for local transactional searches, have a negative correlation with the presence of AI Overviews.

In addition, we see that the presence of sitelinks, which typically indicate that the intent in the query is navigational or a branded term, which is an area where Google is working to reduce the potential for a hallucination.

For marketers who need to detect where AI Overviews could be appearing for their keywords, it will be essential to track the elements in a traditional result that correlate with the presence of an AI Overview. To do this, advanced capabilities such as Data Cube X in BrightEdge allow users to filter results by specific search features and leverage exact match volumes, so marketers can differentiate how long-tail queries, misspellings, and plural/singular versions of keywords may elicit different results.

To begin sensing where AI Overviews will appear, BrightEdge customers should use Data Cube X and filter keywords that contain the following features:

• Featured Snippets
• Questions
• Videos
• Images

SEO Action Items for When AI Overviews Appear:

1. Identify Query Intent: Keywords that are informational are more likely to feature an AI Overview than transactional or navigational queries. Use filtering to identify the questions and top-of-funnel keywords related to your brand.
2. Leverage Traditional Search Features to Pinpoint Where AIO Occurs: BrightEdge users can create keyword groups to track where featured snippets are appearing in search results. These should be monitored and targeted for optimization.
3. Create Content that Addresses the Questions AI Anticipates: Given that AI Overviews are appearing for informational queries and anticipate follow-up questions, marketers need to build content with this framework when targeting keywords where an AI Overview is likely to appear. To do this, content briefs need to address questions related to the keyword or topic being targeted. BrightEdge users can leverage Copilot for Content Advisor to do this faster. It identifies related questions to the topic, scans content currently ranking, and creates briefs and first drafts that address the topics needed to appear in both traditional ranking and AI Overview citations.

 

How are AI Overviews formatted?

Within the various states of AI Overviews, a range of content modules are deployed and tailored to match the user's search intent. These content modules, while diverse in their specific applications, exhibit shared structural characteristics, adapting to the unique nature of each search query.

In describing and categorizing these content modules, we have adhered to the terminologies and descriptions provided by Google, ensuring accuracy and consistency in our analysis. This approach helps delineate how AIO tailors its responses to different search intents, providing valuable insights into the operational mechanics of AIO and its impact on search results.

Unordered Lists

Unordered Lists are utilized in two distinct styles, each serving a specific purpose in presenting information:

Style 1- Standard Bullet Point List: This style is straightforward, featuring a series of bullet points. It is used to present information in a simple, easy-to-scan format, ideal for quick comprehension by users. This format is particularly effective for listing features, benefits, or various options related to a search query.

Style 2- Enhanced List with Breakouts: This style includes additional elements such as expanded descriptions or interactive features within the list. It's designed to provide more detailed information and encourage user interaction. This format is suitable for queries where users might benefit from more comprehensive details or where engagement with the content is crucial.

In addition to unordered lists, ordered lists play a small role in AI Overviews appearing across a wide array of informational queries and spanning various industry types. The manner in which these lists are rendered is closely tied to the specific content they are associated with, reflecting the diversity and adaptability of AIO in presenting structured information.

Key SEO Considerations for Optimizing Unordered and Ordered Lists:

• Utilize Lists in Product Descriptions: When feasible, incorporate  lists into product descriptions. This structured format not only enhances readability but also aligns with AIO’s preference for well-organized content, potentially improving the visibility of your products in search results.
• Craft Concise, Informative Descriptions: Aim to provide succinct yet comprehensive product descriptions. Quality descriptions serve as valuable source material for AIO, enabling it to accurately reflect your products’ features and benefits in its search results.
• Highlight Key Product Features: Pay attention to the characteristics typically used in AIO's ordered lists, such as product specifications, statistics, or unique selling points. Ensuring these elements are clearly presented in your content can aid AI Overviews in accurately representing your products, enhancing their appeal to potential customers.
• Ensure Perks Related to Products Are Marked Up: Perks such as free delivery and stock status should be marked up and present on the product pages. Unordered breakout lists will be offering a deeper shopping experience where users can compare different vendors and their current offers without leaving the search result page.

Product Viewers:

To aid shoppers in understanding different products, AI Overviews offers a couple variations of product viewers which orient the products to what the AI recommends. Based on the type of search, there are a few ways products can be presented in AI Overviews

Standard Product Viewer

A standard product viewer offers multiple images of products with pricing and other pertinent details that align to the query intent

Product Carousels

Product carousels offer less images per product but more shopping detail such as pricing and reviews. These occupy more real estate on a search result page as they offer more products but help the user shop through a wider variety of products.

SEO Implications for Product Viewer Modules:

• Optimize for Trend Relevance: Align your category pages with current trends to increase the likelihood of inclusion in AIO's dynamic displays.
• Incorporate Product Reviews: Facilitate and showcase product reviews on your site to enhance credibility and appeal.
• Category Optimization for AIO Groupings: Adapt your categories to align with AIO's grouping logic, ensuring relevance and visibility.
• Present Clear Pricing Information: Use schema markup to highlight pricing details, enhancing clarity and searchability.
• Monitor Keyword Performance: Keep a close eye on how your key revenue-driving product keywords are performing in AIO results, adapting your strategy in real-time to changes in AIO's presentation formats.

Baseline AI Overview Components

AI Overviews, regardless of the specific state or module in use, incorporates a set of standard features in its results. These features are consistent across various queries and provide a structured approach to presenting information.

Answer Box:

• Positioned at the top of the search result.
• The content within this box is AI-generated and tailored to the query. It may include text and, in some cases, graphics.
• The sources for the information are accessible by clicking on a chevron icon within the box.
• Google includes a disclaimer regarding the potential variability in information quality.

Source Panel:

• Located to the right of, or just below the Answer Box.
• Displays the primary sources (up to three initially visible, with ten available upon scrolling) that the AIO uses for generating the Answer Box and other components, such as ordered lists.
• The sources for the information are accessible by clicking on a chevron icon within the box.

Special Cases

As Google progresses in refining AI Overviews, specialized modules are appearing for small sets of queries. Specific elements, tailored to distinct query types, are being introduced. These elements are designed to either integrate cohesively with or complement the existing search modules.

Pro/Con

For a small number of queries, we detect pro/con boxes that are unordered lists that weigh pros and cons of products.

Combined Shopping Offer Cards

For certain e-commerce queries, BrightEdge Generative Parser™ detects shopper offer cars which offer pricing, as well as offers such as free delivery in a compact view. While it remains uncertain if these are sourced from the Schema "offer" tag, it's evident that signals aligning with what the offer card displays could enhance the likelihood of appearing in these results:

Hero Images

Currently, when activated, these elements consistently appear at the bottom of the answer box. Notably, all images appear to be sourced. To enhance eligibility for display in these sections, it is crucial to ensure that images are optimized with relevant tags and are optimally compressed. This strategy underscores the importance of image optimization in facilitating effective and relevant search result displays:

 

How do AI Overviews (AIO) overlap with traditional results

AI Overviews are designed to provide additional context to queries that users may find helpful. As a result, many of the citations used by AI Overviews are different from those that rank in traditional results. This is because AI Overviews aim to answer the next question a user might have after their initial query, without prompting the user to ask that second question.

Across all keywords, we see that the URLs cited in AI Overviews are not ranking in traditional results 66% of the time.

The focus on content that isn’t currently ranking is a result of Google focusing on making AI Overviews as helpful as possible. Instead of simply repeating the information users could find within the rankings, the AI Overview anticipates the next question users would have and cites content that likely wouldn’t fit the immediate keyword ranking.

For marketers, this presents an opportunity to address multiple questions with multiple URLs in a single search result.

What Domains Are Sourced the Most?

As indicated, we are seeing our data skew to healthcare. Despite concerns about citation trustworthiness in the news, at the macro level, the share of citations supports the notion that the more trusted and authoritative the website is, the more likely it is to be cited.

Within each industry, the websites cited to support AI Overviews also vary. We also see that within industries, there are leaders that occupy a greater share of citations than others.

E-Commerce

• E-commerce features a wider dispersion of citations than other categories, with the leader having 27% of the citations (the average is 35%).
• Cited sources are a combination of retailers and publishers, which coincides with the modules used in AI Overviews (editorial content as well as product modules).
• For marketers, it is important to understand the query intent (and optimize for the modules that are likely to appear) as well as what questions the AI Overview will anticipate the user will have.

Rank Overlap for E-Commerce

• E-commerce queries are close to what we see in industries as a whole, where roughly 75% of citations are not ranking in traditional results.
• Many of the AI Overviews in e-commerce include multiple modules (a paragraph as well as a product carousel), which provides more opportunities for a wider range of websites to rank.
• For marketers targeting e-commerce, where an AI Overview is likely to appear for 23% of queries, it is important to understand both the transactional aspect of the overview (the product carousel) and the informational element, and build a content strategy to target both.

​B2B Technology

• Techtarget features a large resource library in multiple formats specific to B2B technology.
• The distribution of share of citations in B2B Tech is similar to what we see at the macro level.
• There is a combination of Fortune 500 and industry-specific sites that occupy the most citations, suggesting marketers representing sites of all sizes have an opportunity to be cited.

B2B Rank Overlap

• In B2B Tech, 20% of page 1 rankings are used in the AIO, which is greater than the average.
• For marketers targeting B2B technology queries, blogs and long-form content that can help answer adjacent questions related to the query could help ensure domains can win in both AI citations and traditional rankings.

Education

• Citations in AIO for education are led by a site dedicated to jobs, which may indicate that AI infers more about careers for educational queries than registering for classes.
• A range of authoritative .edu sites are cited heavily. Education features an AI Overview 14% of the time, which is slightly above average.
• LinkedIn has a share of citations, indicating that user-generated content could be effective here.

Education Rank Overlap

• Education AI Overviews have less than average overlap, with more than 90% of citations not ranking in traditional results.
• The presence of sites like Indeed suggests that AI Overviews infer the second and third searches a user might make in education, necessitating different citations.
• For marketers in education, social content (e.g., LinkedIn) and editorial content that focuses on additional questions should be integral to content strategies to win in AI Overviews.

Entertainment

• Entertainment draws on a variety of resources from trusted authoritative sites such as IMDB to reddit and medium which feature user generated content.
• Entertainment as a category has an AI Overview on approximately 14% of keywords tracked in BrightEdge Generative Parser™
• For marketers targeting entertainment, ensure informational content features schema markup that could help AI understand where content could be used in formats such as unordered lists.

Rank overlap for Entertainment

• Less than 10% of AI Overviews for entertainment are cited by a ranking in the top 10.
• Based on keyword analysis, it was discovered that when keywords feature "best," the likelihood of an AI Overview increases by 40%.
• For marketers in entertainment, monitor queries that include "best" and any that have featured snippets. These queries are far less likely to rank a page already ranking in organic results than other queries.

Finance

• Finance queries are similar to what we expect to see in traditional ranking results, which aligns with the higher trust required for YMYL (Your Money or Your Life) queries.
• The distribution is slightly more concentrated than the average, with Investopedia winning 42% of citations (the overall average for the top position is 35%).
• For marketers targeting finance queries, E-E-A-T (Experience, Expertise, Authoritativeness, and Trustworthiness) is essential, and associations with leading finance sites, as mentioned above, should help reinforce trust in your site.

Rank overlap for Finance

• We see a trend similar to the overall benchmarks where 84% of queries do not rank in the top 10. Despite this, many of the top-cited websites are also ranking highly in traditional search, suggesting that AI Overviews may look to answer the next question but still focus on content from similar sites.
• Overall, Finance is a low AI Overview category, with only 5% of keywords showing an AI Overview.
• For marketers targeting Finance, continue to flag keywords where featured snippets are appearing, build relationships with the top sourced sites, and focus on the second and third questions that a user may have after their initial query.

Healthcare

• Healthcare is the industry where the most AI Overview deployments take place. Presently, 63% of healthcare queries have an AI Overview. This data is current as of 6/12/2024. Contrary to reports about AI Overviews relying on untrustworthy sources, the data shows that the websites cited for AI Overviews are very authoritative.
• We see almost double the concentration of citations for healthcare sites compared to the average. NIH.gov has 60% of the share of citations for healthcare (across all industries, the top domain is 35%).
• For marketers in healthcare, expect AI Overviews to appear for informational queries. Identify areas where you can partner with trusted sites and focus on the second and third questions related to queries that have a featured snippet.

Rank Overlap for Healthcare

• We see the greatest concentration of page 1 rankings for healthcare, with 30% of cited URLs also ranking on page 1.
• The uptick in pages not ranking for the result suggests AI Overviews are going further to answer the second and third question
• For marketers in healthcare, the data suggests traditional rankings are important indicators for the likelihood of appearing in AI Overviews. Once this is achieved, focus on the related questions about the topic to ensure content is cited.

Insurance

• AI Overviews appear for 18% of insurance queries. Based on the share of citations, it appears to be treated similarly to YMYL (Your Money or Your Life) with a set of trusted sites holding the largest shares (nearly half of AIO citations are from dol.gov).
• Despite this, there do appear to be opportunities for smaller sites to build citations (e.g., wearedevelopers.com, timechamp.io).
• For marketers, regardless of the site size, watch where Featured Snippets are appearing and build answers to questions around the keyword in your content.

Rank Overlap for Insurance

• Insurance is trending slightly higher than the average where we are seeing only 75% of results not ranking in traditional organic rankings
• For marketers, insurance underscores the importance of topic clusters where your primary ranking page can target a keyword, but then secondary pages can focus on the second and third questions a user may ask that AI Overviews will anticipate.

How do AI Overviews (AIO) Differ from Search Generative Experiences (SGE)?

AI Overviews are the refinement of a series of tests and deployments that Google conducted for nearly a year with Search Generative Experiences. As a result of these tests, there are refined conditions that have strategic implications.

AI Overviews do not offer follow up questions

With SGE, Google prompted users to ask a followup which would initiate a conversation with the AI:

AI Overviews have scaled back this functionality, which may indicate that Google does not want to cannibalize engagement from the rest of the search result page.

Today, AI Overviews are appearing in significantly fewer queries, implying that Google has fine-tuned where AI Overviews are most helpful, thus reducing the need for follow-ups. Additionally, many anticipated questions are addressed directly in the AI Overview, negating the need for user prompts. For example, in the query “what are training shoes,” the AIO anticipates a question about the differences between training shoes and running shoes and answers it immediately, without requiring further interaction from the user. It also cites sources for answers to anticipated questions.

AI Overviews are 20% Smaller

Across the board, due to changes in how modules are presented and how content is consolidated, AI Overviews take up less space on the screen than SGE did. This means that other aspects of the search result page, such as traditional organic results and sponsored ads, are less likely to be crowded out when they appear. On average, the BrightEdge Generative Parser™ has detected that these are 20% smaller than their SGE counterparts. The difference in screen real estate is noticeable.

For marketers, this means that AI Overviews are less likely to diminish clicks from other elements of the search result page. However, it also means that citations and the opportunity to gain real estate within the AIO for a keyword are likely to be more competitive, as there is less space to compete in.

While SGE allowed users to respond with a vote on whether they liked a result or not, AI Overviews allow users to provide specific feedback on the result to help Google fine-tune it. This is a particularly important feature as the AI Overview rollout has faced some challenges with accuracy. Additionally, it gives marketers an opportunity to report inaccuracies on their own branded queries. For example, if a query contained inaccurate information about Macs, the digital marketer could provide that feedback directly to Google.

For marketers, this is an opportunity to include brand accuracy into workflows. Monitor brand queries for the presence of AI Overviews and use the Feedback Module to address any inaccuracies or editorial content that isn’t accurate.

AI Overviews are using more space efficient modules to display products

Ecommerce is a category where AI Overviews are appearing more frequently than on average (20% instead of 10%). Throughout the SGE testing period, the most common way to display products was through a module called a Product Viewer, which featured the product standalone with content generated by Google.

With the launch of AI overviews, these results were decline in visibility as they became more in line with carousel results:

We saw an overall decrease in the presence of product viewer formats when AI Overviews were launched, as Google reduced the overall deployment of AIO. Since that time, however, we are seeing a steady rise of carousel groups, which allow for more products to appear in less space than a product viewer.

For marketers in ecommerce, this means it is critical to ensure your product content has the appropriate schema markup and context, including reviews, images, pricing, and supplementary information such as inventory counts, so AI can present them alongside competing products.

Conclusion

A lot of the codebase and principles uncovered during the lab phase of SGE carry through to AI Overviews. However, with AI Overviews being more selective in where they appear and serving a more defined purpose, marketers need to ensure sensing and reacting to them is part of their content strategy.

With BrightEdge, marketers can leverage Data Cube X to filter on features such as Featured Snippets to pinpoint which of their keywords are likely to trigger an AI Overview. These are then integrated into a workflow to ensure they are optimized for both traditional rankings and AI Overview citations.

The data across these industries demonstrate how AI Overviews do not rely on the same ranking methods as traditional algorithms. In many cases, AI Overviews are sensing and anticipating questions that may also be asked. To build content that addresses both current and anticipated questions, marketers can leverage BrightEdge Copilot for Content Advisor. This tool uses AI to gather all the concepts and questions related to a topic to generate briefs and first drafts that will win in traditional search and be ideal for AI Overview citations.